[c-nsp] RPF on 6500

[Jared Mauch]

On Sun, May 22, 2005 at 10:14:32AM -0700, Tim Stevenson wrote:
> At 06:08 AM 5/22/2005, RawCode gushed:
> >Is there a performance difference between the two methods (rx | any)
> >of turning on RPF on a 6500?
> 
> No, same performance impact (ie, none). Do note that you can't mix/match 
> modes on different i/fs, all i/fs use the last configured mode, strict or 
> loose.

	Keep in mind that this is the case for the sup2, the sup1
it's done in SW, the sup3(720) shouldn't have a problem
with this.

	Watch out, as we found this limitation in operation the hard
way, also it reduces the table from 256k to 128k in the sup2,
so it's not really recommened at all on this platform if you're
doing full bgp, as there are more than 128k routes in the table..

	This can cause funky things like your input queue being wedged
because the sup is at 100% cpu and the mfsc is sitting idle, but in the
case of interface stats (snmp polling) those are kept on the sup,
so will cause fun IPC issues btw the sup <-> mfsc and your device
to go down since there is no internal timeout mechanisim for packets
in the input "queue".

	- jared

-- 
Jared Mauch  | pgp key available via finger from jared at puck.nether.net
clue++;      | http://puck.nether.net/~jared/  My statements are only mine.